Closed loop pulsating heat pipes Part B: visualization and semi-empirical modeling

(1)

Closed loop pulsating heat pipes

Part B: visualization and semi-empirical modeling

Sameer Khandekar

a,*

, Piyanun Charoensawan

b

,

Manfred Groll

a

, Pradit Terdtoon

b

a

Institut f€uur Kernenergetik und Energiesysteme (IKE), Universit€aat Stuttgart, Pfaﬀenwaldring 31, 70569 Stuttgart, Germany

b

Department of Mechanical Engineering, Chiang Mai University, Chiang Mai 50200, Thailand Received 4 April 2003; accepted 15 May 2003

Abstract

Pulsating heat pipes have received growing attention from experimental and theoretical researchers in recent times. Behind its constructional simplicity lie the intriguingly complex thermo-hydrodynamic op-erational characteristics. Part A of this paper has presented the thermal performance results of a fairly large matrix of closed loop pulsating heat pipes. This paper, which is an extension of the previous work, ﬁrst presents some more visualization results to highlight the complexities involved in mathematical formulation of the modeling problem. The phenomenological trends recorded in the visualization set-up are fully inline with the previous quantitative data of Part A. Critical review of the existing modeling approaches to CLPHPs is presented in the wake of these results. A detailed discussion follows on the important issues involved in the mathematical modeling of these devices. Then, semi-empirical correlations based on non-dimensional numbers of interest for predicting the thermal performance of CLPHPs are presented. Al-though there are limitations of the models presented herein, modeling by non-dimensional numbers seems to be most promising as compared to other existing techniques.

Keywords: Pulsating heat pipe; Semi-empirical correlation

*_{Corresponding author. Tel.: +49-711-685-2142; fax: +49-711-685-2010.}

(2)

Nomenclature

Bo Bond Number

Cp constant pressure speciﬁc heat, J/kg K

D tube diameter, m E€oo E€ootv€oos Number f friction factor

g gravitational acceleration, m/s2

h heat transfer coeﬃcient, W/m2_K

hfg latent heat, J/kg Ja Jakob number Ka Karman number k thermal conductivity, W/m K L length, m _ m

m mass ﬂow rate, kg/s N number of turns P pressure, N/m2

Pr Prandtl number _

Q

Q heat transfer rate, W _q

q heat ﬂux, W/m2

Re Reynolds number T temperature, C or K t _{characteristic velocity m/s}

x _{two-phase mass quality}

Greek symbols

b inclination angle from horizontal axis, radian l dynamic viscosity, Pa s q density, kg/m3 r surface tension, N/m Subscripts a adiabatic section c condenser section CB convective boiling crit critical value e evaporator section exp experiment i inner l liquid NB nucleate boiling o outer pre predicted

sat saturation condition

(3)

1. Introduction

A modest number of studies, mostly qualitative and some quantitative, have been performed so far to understand the thermo-hydrodynamics of pulsating heat pipes (e.g. [1–4]). Various oper-ational characteristics, parametric experiments and phenomenological studies have indicated that this family of heat pipes is extremely complex from many aspects of thermo-fluidic sciences. Although important milestones have been achieved in specific areas, a comprehensive under-standing is still lacking. Parallel research is also underway in the direction of mathematical modeling of pulsating heat pipes. Looking at the available literature it can be concluded that success in this direction is only marginal. Many simplified approaches have been attempted which may be categorized according to the simplification scheme adopted. These may be summarized as: Type 1: comparing PHP action to equivalent single spring-mass-damper system [5], Type II: ki-nematic analysis by comparison with a multiple spring-mass-damper system [6], Type III: ap-plying conservation equations of mass, momentum and energy to specified slug–plug control volume [7,8], Type IV: analysis highlighting the existence of chaos under some operating condi-tions [9] and Type V: modeling based on Artificial Neural Networks, wherein an ANN archi-tecture is trained to predict performance with the available experimental database [10]. Extreme simplification has been adopted in all the above approaches and the results have only limited validity and contribution in overall understanding of the device, not to mention in performance prediction and optimization [11].

In the present work, firstly the general strategy of mathematical modeling of closed loop pul-sating heat pipes is discussed taking into consideration the experimental evidences available so far. Then a semi-empirical modeling scheme incorporating non-dimensional numbers of interest is presented. The non-dimensional numbers affecting the thermo-fluidic behavior are identified and their direction of influence is explained. The experimental data presented in Part A of this paper is used to construct the model [12]. In addition, apart from drawing information from other visu-alization studies, glass tube transparent CLPHP structures are fabricated to augment the quan-titative results. The visualization trends clearly indicate the strong dependence of the thermal performance of CLPHPs on the flow patterns existing in the device. Based on the results of the present study and the discussion of the issues related to mathematical modeling of pulsating heat pipes provided before the background of the visualization results, it seems that semi-empirical approaches may prove to be the most promising in modeling CLPHPs amongst all the existing techniques applied so far.

2. Experimental set-up (for ﬂow visualization)

In Part A of this paper, the experimental set-up and plan was outlined for the quantitative measurements performed on CLPHPs made of copper tubes [12]. To augment the understanding of these quantitative results, visualization experiments were planned. These were deemed neces-sary in view of the already existing hypothesis that the thermal performance (e.g. the overall thermal resistance) is largely dependent on the two-phase ﬂow patterns existing in the device. The visualization experiments have indeed proven this hypothesis as will be demonstrated in the results.

(4)

Most of the peripheral equipment in the visualization experimental set-up was the same as already described in Part A of this paper [12]. The visualization experimental set-up details are shown in Fig. 1. Transparent CLPHPs were completely made of Pyrex glass tubes with Di¼ 2

mm. The evaporator section was heated by silicone oil, the inlet temperature of which was always maintained at 80 C. The mass ﬂow rate was adjusted to maintain near isothermal boundary conditions. The condenser section was cooled by distilled water always maintained at 20C. From the outlet temperature and mass ﬂow rate of the coolant, the heat transfer could be calculated. A total of three set-ups were built, i.e. CLPHP with Le¼ 50 mm with 10 turns, Le¼ 50 mm with 28

turns and Le¼ 150 mm with 11 turns (Le¼ Lc¼ La in all cases, as in Part A). All CLPHPs were

tested with R-123 as the working fluid at a fixed filling ratio of 50% and inclination angles of 0, 30, 50, 70 and 90 from the horizontal axis. After a quasi steady state was reached at a given orientation, continuous movies were recorded by video cameras (Sony, CCD-TR618E) while photographs were taken at specified times by digital still cameras (Nikon Coolpix 5700, Sony, DSC-S75).

3. Results and discussion

3.1. CLPHP thermo-hydrodynamics

In order to formulate the mathematical models describing the complex operation of CLPHPs, a thorough understanding of the internal thermo-hydrodynamic phenomena is an essential pre-requisite. A brief look at the operating mechanism of the device reveals that there is multitude of influence parameters affecting the thermal performance. It is rare to find a combination of such events and mechanisms, like bubble nucleation and collapse, bubble agglomeration and pumping action, pressure/temperature perturbations, flow regime changes, dynamic instabilities, metastable non-equilibrium conditions, flooding or bridging etc., all together contributing to-wards the thermal performance of a device. Therefore, the issue of mathematical modeling

(5)

of CLPHPs has to be dealt with only in the background of the established experimental evi-dences.

Although the internal fluid flow pattern has been classified, in general, as capillary slug flow, transition to semi-annular/annular flow has been observed earlier [4] as well as in the present experiments. The slug-annular transition depends not only on the heat input but also on the geometrical constructional features and inclination angle of the device operation. The length of tube sections in the condenser and evaporator determines the flux at which heat is being rejected and fed in. Entrainment due to Helmholtz type instabilities and subsequent ÔbridgingÕ/ÔfloodingÕ phenomena and other forms of dynamic instabilities are the limiting factors of the slug-annular transition. For a given geometry, this transition usually occurs with increasing input heat flux. Even when the flow pattern is predominantly slug flow, various bubble patterns have been ob-served. Not only various bubble agglomeration and breaking patterns are evident, different flow patterns were also present in conjunction (as cited in [12]). The phenomenological trends for maximum heat throughput, as observed in the visualization set-up, were exactly inline with the quantitative results already discussed in Part A, i.e. the performance independence with orien-tation is affected by the number of turns, and that the effect could be clearly separated into two cases by using a certain critical value of number of turns (Ncrit) (refer Fig. 2). When N is less than a

certain Ncrit, the CLPHP cannot satisfactorily operate in the horizontal orientation and vice versa

(for details, refer Part A, i.e. [12]). The visualization results showed that the maximum heat transported in different subsections (Zones A1, A2, B1 and B2) of Fig. 2 is clearly dependent on different internal flow patterns, as indicated by the evaporator section images shown therein. Near to horizontal position, no annular flow is observed and there is very little bulk movement. Slugs/ plugs only vibrate with high frequency and low amplitude about a mean position. All bulk movement tends to stop (standstill condition) for a while and then restarts in a cyclic manner. Large vapor bubbles encompass the evaporator bends with hardly any bulk movement. This stopping time is an order of magnitude greater in case when N < Ncritas compared to the case of

N > Ncrit. For situations that result in predominantly slug ﬂow conditions (i.e. b < 30), ﬂow

direction rapidly changes and is rather ÔchaoticÕ. When the device is tilted towards the vertical, transition to semi-annular/annular flow is observed in both the cases (Zone B1 and B2). When N < Ncrit annular flow tends to develop but reverse transition to slug flow occurs (bridging/

ﬂooding) before reaching the condenser section. When N > Ncrit, annular ﬂow is seen until well

inside the condenser section. Bulk flow also tends to take a fixed direction and alternate tubes are then hot and cold as seen in Fig. 3. The heat throughput is the best in such cases; it follows from the fact that increased tendency of annular flow (convective boiling) in the evaporator increases the net heat transfer coefficient.

Form the study it may be further concluded, partly by conjecture and partly by visual exper-iments that dynamic two-phase flow instabilities are an inherent part of CLPHP operation and metastable non-equilibrium flow conditions will exist in the tubes. Flashing of the vapor is im-minent as the fluid pressure falls during the flow. Instantaneous subcooling and superheating are also expected as heat/mass transfer has finite inertia. Thus, the existing mathematical models as outlined in the earlier section (Type I, II and III), that are based on ideal slug flow assumption with no bubble agglomeration/breakage and without consideration of any flow regime change associated with geometry, global orientation, number of turns, heat input etc., seem to be com-pletely missing the thermo-hydrodynamic physics of the device. The Type IV models only point

(6)

out that Ômathematical chaosÕ exists in CLPHPs but cannot predict the device thermal perfor-mance. Lastly, the ANN based model, which is indeed a ÔBlack BoxÕ approach, can be eﬀective only if a large amount of reliable data, covering all the aspects of CLPHP thermo-hydrodynamics, is available. It can therefore be concluded that other modeling strategies have to be adopted for performance prediction.

3.2. Critical issues in modeling

One critical issue, i.e. the existence of bubble agglomeration and breaking including various two-phase flow patterns, has already been pointed out in the previous sub-section. This, in itself makes analysis quite difficult. In addition, from the normal operation of the device, it is clear that flow boiling occurs in CLPHPs. The type and magnitude of internal ÔflowÕ is a direct consequence of the applied heat flux, geometry and the inclination angle of operation. The input heat flux depends on N , Le and Bo (which is non-dimensional Di). For example, if Bo is larger than a

Fig. 2. Phenomenological trend for heat throughput in experimental data reported in [12] are shown above. This trend is the result of various ﬂow patterns occurring in diﬀerent subsections, as depicted. The images are taken in the evaporator section of the visualization set-ups.

(7)

specified critical Bo, then all the working fluid will tend to settle down by gravity and the device will no longer be a pulsating heat pipe. Instead, it will function as an interconnected array of two-phase gravity assisted thermosyphons, with pool boiling dynamics primarily governing the per-formance. Thus, for a given geometry and fill charge, the input heat flux and the ÔflowÕ that results thereof leads not only to different two-phase flow patterns, as noted earlier, but also to different heat transfer governing mechanisms.

On one extreme with very low input heat fluxes, Ônucleate pool boilingÕ or Ôfree convection nucleate boilingÕ may be the only dominant phenomena. A little increase in the heat flux will slowly give rise to Ônucleate flow boilingÕ. On the other extreme, continuous increase of input heat flux will result in a condition where nucleate boiling may be completely suppressed and Ôcon-vective flow boilingÕ becomes the dominant phenomenon (e.g. if annular flow quickly develops in the evaporator and heat transfer takes place through the liquid film–vapor interface). Thus, different input heat fluxes to the same device give rise to a range of heat transfer scenarios.

In nucleate boiling, the heat transfer coefficient is chiefly dependent upon the heat flux and practically not at all upon the flow velocity. On the contrary, in convective boiling, the heat transfer coefficient is primarily influenced by the velocity of flow or by the mass flux _mm, but on the other hand is scarcely influenced by the heat flux [13]. Vapor mass quality is an additional in-dependent variable for convective boiling. These facts give rise to the hypothesis that the time averaged effective heat transfer coefficient obtained in the evaporator U-turns of CLPHPs is a superposition of the respective heat transfers obtained by free nucleate boiling and convective flow boiling.

ðheffÞCLPHP ¼ S hNBþ F hCB ð1Þ

Fig. 3. Typical operation in Zone B1, vertical position, is seen here. The images are taken in the adiabatic zone of the visualization set-up. Semi-annular/annular ﬂow exists in alternating tubes.

(8)

The quantities S and F are essentially adjustment parameters (also referred to as ÔsuppressionÕ or ÔenhancementÕ factors [14]) to take into account the relative importance of nucleate pool boiling and convective ﬂow boiling mechanisms. Such a hypothesis is also adopted frequently in modeling usual convective ﬂow boiling in open systems. In open systems, the basic empirical correlations for modeling hNB are of the form,

hNB/ C1 _qqn ð2Þ

More detailed and well known models prescribed for hNB by Rohsenow [15] and Forster

and Zuber [16] are based on the assumption that the process of bubble growth and release in-duces motions of the surrounding liquid that facilitates convective transport of the heat from the adjacent surface. This hypothesis leads to the following generic form of models describing hNB, ðNuÞ_NB¼ hNB L kliq ¼ C ðRe_Þn ðPrliqÞm ð3Þ

In this equation the characteristic length and velocity scales (L_{and Re}_{) are appropriately selected}

as per the situation (e.g. bubble departure diameter and vapor superﬁcial velocity). It is interesting to note that by such an appropriate substitution, these Nusselt–Reynolds number correlations can be reduced to forms quite similar to Eq. (2) above. The success of these equations lies in the fact that it has been possible to explicitly deduce characteristic length and velocity scales form the thermo-mechanical boundary conditions of the problem.

For modeling hCB in open systems, additional effects of the mass flux _mm and two phase flow

quality x must be included so that the equation takes the form,

hCB/ C1 _qqn ð _mmÞs UðxÞ ð4Þ

where the quantity C1 is a strong function of the heated surface and the properties of the boiling

liquid. In this case too, the characteristic liquid velocity scale is known a priori as the mass ﬂux and geometry are speciﬁed in the open system.

Major hurdles in practical realization of Eq. (1) for design purposes in case of CLPHPs are as follows:

• The system is closed.

• The characteristic velocity scale is not explicitly known. It strongly depends on the input heat ﬂux and other geometrical constructional features of the device. In addition, the Ôﬂow velocityÕ has an oscillating character.

• In general, there is not enough quantitative flow and heat transfer data available on two-phase open systems in mini-micro channels. More data and subsequent understanding is required on flow instabilities, pulsating and oscillating flows in mini-micro capillary tube systems and re-lated issues. Data on CLPHPs is still more scarce.

Thus, it may be concluded that there are several complex inter-related issues that distinguish mathematical modeling of CLPHPs from conventional modeling of ﬂow boiling in open systems.

(9)

3.3. Non-dimensional groups of interest

In order to find heat transfer correlations having broad applicability range, it is generally convenient to state the decisive thermo-physical properties for heat transfer as non-dimensional quantities. This proposal is a direct and natural consequence of foregoing discussion on the critical issues involved in modeling of CLPHPs. In addition, an exponential form of the corre-lation is preferable since such equations are well proven for heat transfer phenomena. From the quantitative investigations of CLPHPs presented in Part A and in conjunction with the internal flow patterns described herein, it can be concluded that the influence parameters consist of the inclination angle (b), the thermo-physical properties of working fluid, the tube inner diameter (Di),

the number of turns (N ) and the evaporator length (Le) as deﬁned by the following equation,

_q

q¼ QQ_

p Di N 2Le

!

¼ Uðb; geometry; thermo-physical properties of working fluidÞ ð5Þ Since the experiments were conducted at a fixed filling ratio of 50% of the working fluid, this parameter does not appear in the above equation. For a more generalized case, the effect of filling ratio also has to be included. The identification of the important non-dimensional groups which affect the heat transfer characteristics of the device and the proper definition of the function U is needed for practical exploitation of Eq. (5).

Before proceeding further, it should be pointed out that, in general, experiments on CLPHPs may be performed in two ways (i) either by controlling the input heat ﬂux and the condenser temperature; in this case the evaporator temperature is the dependent parameter, or (ii) by controlling the evaporator and condenser temperatures (DTec

sat ); the heat throughput then

be-comes the dependent variable. The experiments reported in Part A were done with the latter strategy. In doing so, the maximum temperature difference occurring in the device was fixed and this, in turn, gave the corresponding saturation pressure difference (DPec

sat ).

The equations stated in the previous section already provide a broad hint of the important non-dimensional quantities of interest. It is indeed obvious that some representative form of flow Reynolds number affects the heat transfer. As noted earlier, it is essential to find the characteristic velocity scale as per the boundary conditions of the problem. Since the experiments reported in Part A were done by maintaining the evaporator and condenser temperature constant (i.e. DTec sat

constant), the characteristic Reynolds number may be calculated as follows: Reliq¼ q_liq t_D i l_liq ! ð6Þ and since the pressure drop in pipe ﬂows is given by

ðDP Þ_liq/f ðqliq ðt

_Þ2

Þ ðDi=LeffÞ

ð7Þ substituting Eq. (7) into Eq. (6) deﬁnes a dimensional group, sometimes referred to as Karman number in the context of pipe ﬂows:

(10)

Kaliq¼ f Re2_liq¼ q_liq ðDP Þ_liq D2 i l2 liq Leff where Leff ¼ 0:5ðLeþ LcÞ þ La ð8Þ

The fact that Karman number is calculated for the liquid phase (and not for an equivalent ho-mogeneous two-phase mixture) is based on the assumption that out of the total ﬂuid losses, consisting of ðDP Þ_liq and ðDP Þ_vap, the former is an order of magnitude higher than the latter. Therefore, in the above Eq. (8),ðDP Þ_liq ðDP Þec_sat . Thus, the Karman number gives an appropriate velocity scale for CLPHP modeling.

The next numbers of interest, which do not need explanation, are the liquid Prandtl number and the Jakob number, deﬁned respectively as,

Prliq¼ Cp;liq lliq kliq ð9Þ Ja¼ hfg

Cp;liq ðDT Þecsat

ð10Þ the thermo-physical properties being calculated at ðTeþ TcÞ=2. The liquid Prandtl number scales

the single phase convective eﬀect on heat transfer while the Jakob number highlights the relative importance of sensible and latent portions of heat transfer in a CLPHP. In addition, from the generic trends of the experimental data reported in Part A, an exponential function was found to be the best representation of the eﬀect of inclination angle. Thus, Eq. (5) can be written as,

_q

q¼ QQ_

p Di N ð2LeÞ

!

¼ C1 ðexpðbÞÞp Kaq_liq Pr_liqr Jas ð11Þ

In the above equation the constant C1 has the dimension of W/m2.

The Bond number (or alternatively the E€ootv€oos number) is the ratio of surface tension and gravity forces and deﬁned as follows,

Bo¼ Di

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi gðqliq qvapÞ=r

q

¼pffiffiffiffiffiffiE€oo ð12Þ

If the Bond number exceeds a particular critically value (Bocrit 2) stable liquid slugs will not

form and the device will not function as a pulsating heat pipe, as explained in Part A of this paper [12]. For Bo > 2 (for example standard closed two-phase thermosyphons), the heat transfer limitation comes from nucleate pool boiling and counter-current ﬂow limitations [17–19]. Based on these facts it may be said, although without existence of evidence, in favor or otherwise, that Eq. (11) is only valid for situations where Bo 6 Bocrit.

An attempt was made to correlate the entire data sets (a total of 248 data) of the experimental matrix reported in Part A [12]. This resulted in a correlation given by Eq. (13) below for which the ﬁt is depicted graphically in Fig. 4. Standard least square curve ﬁtting technique coupled with Gauss elimination method was adopted resulting in an overall drift in predictions within 30%.

_q

q¼ QQ_ pDi N 2Le

!

(11)

Eq. (13) essentially determines the maximum heat transfer achievable for a given CLPHP (with filling ratio ¼ 50%) which is imposed to a specified temperature difference DTec

sat between the

evaporator and the condenser. Alternatively, if heat ﬂux and Tc is known, then an iterative

so-lution by guessing Te can be employed.

4. Summary and conclusions

The following facts summarize the essential aspects of this study:

• Closed loop pulsating heat pipes are complex heat transfer systems with a very strong thermo-hydrodynamic coupling governing the thermal performance.

• Different heat input to these devices give rise to different flow patterns inside the tubes. This in turn is responsible for various heat transfer characteristics. The study strongly indicates that design of these devices should aim at thermo-mechanical boundary conditions which result in convective flow boiling conditions in the evaporator leading to higher local heat transfer co-efficients.

• The inclination operating angle changes the internal ﬂow patterns thereby resulting in diﬀerent performance levels.

• The available models in the literature do not truly represent the thermo-hydrodynamics of the CLPHPs. In addition, models applicable for open systems are not directly applicable for closed CLPHPs. Therefore a semi-empirical correlation has been developed to ﬁt the available data based on non-dimensional numbers of interest. The Karman number as deﬁned in this paper represents a suitable velocity scale for CLPHPs.

• In the wake of the critical issues regarding modeling of CLPHPs discussed in this paper, a semi-empirical approach seems to be quite satisfactory. For a more fundamental modeling approach,

(12)

an additional reliable data base is needed in congruence to the wider research in mini/micro scale boiling heat transfer in open systems.

Acknowledgements

This research work was done jointly by Faculty of Engineering, Chiang Mai University, Thailand and Institut f€uur Kernenergetik und Energiesysteme (IKE), Universit€aat Stuttgart, Ger-many under the auspices of Royal Golden Jubilee Scholarship of the Thailand Research Fund (under contract no.: 1.M.CM/43/A.1) and Deutscher Akademischer Austauschdienst (DAAD). The work was also partly supported by Deutsche Forschungsgemeinschaft (DFG) under Grant GR-412/33-1.

References

[1] H. Akachi, F. Polaassek, Thermal control of IGBT modules in traction drives by pulsating heat pipes, in: Proceedings of 10th International Heat Pipe Conference, vol. 3, Stuttgart, 1997, pp. 8–12.

[2] M. Hosoda, S. Nishio, R. Shirakashi, Study of meandering closed-loop heat-transport device (vapour-plug propagation phenomena), JSME Int. J. B 42/4 (1999) 737–744.

[3] S. Khandekar, M. Schneider, P. Sch€aafer, R. Kulenovic, M. Groll, Thermoﬂuiddynamic study of ﬂat plate closed loop pulsating heat pipes, in: Microscale Thermophysical Engineering, vol. 6(4), Taylor and Francis, ISSN 1089-3954, 2002, pp. 303–318.

[4] M. Groll, S. Khandekar, Pulsating heat pipes: progress and prospects, in: Proceedings of 3rd International Conference on Energy and Environment, Shanghai, 2003, in press.

[5] Z. Zuo, M. North, Miniature High Heat Flux Heat Pipes for Cooling of Electronics, Thermacore Inc., 2000, Available atwww.thermacore.com.

[6] T. Wong, B. Tong, S. Lim, K. Ooi, Theoretical modeling of pulsating heat pipe, in: Proceedings of 11th Interantional Heat Pipe Conference, Tokyo, 1999, pp. 159–163.

[7] G. Swanepoel, A. Taylor, R. Dobson, Theoretical modeling of pulsating heat pipes, in: Proceedings of 6th International Heat Pipe Symposium, Chiang Mai, 2000.

[8] M. Shaﬁi, A. Faghri, Y. Zhang, Thermal modeling of unlooped and looped pulsating heat pipes, J. Heat Transfer, ASME 123 (2001) 1159–1172.

[9] S. Maezawa, S. Nakajima, K. Gi, H. Akachi, Experimental study on chaotic behavior of thermohydraulic oscillation in oscillating thermosyphon, Heat Pipe Technology (Proceedings of 5th International Heat Pipe Symposium), Pergamon Elsevier, ISBN 0-08-042842-8, Melbourn, 1996, pp. 131–137.

[10] S. Khandekar, X. Cui, M. Groll, Thermal performance modeling of pulsating heat pipes by artiﬁcial neural network, in: Proceedings of 12th International Heat Pipe Conference, Moscow, 2002, pp. 215–219.

[11] S. Khandekar, M. Schneider, M. Groll, Mathematical modeling of pulsating heat pipes: state of the art and future challenges, in: Proceedings of 5th ASME/ISHMT Joint International Heat and Mass Transfer Conference, Tata McGraw Hill, ISBN-0-07-047443-5, Kolkata, 2002, pp. 856–862.

[12] P. Charoensawan, S. Khandekar, M. Groll, P. Terdtoon, Closed loop pulsating heat pipes––Part A: Parametric experimental investigations, Applied Thermal Engineering, Elsevier Science, ISSN 1359-4311, in press, 2003. [13] K. Stephan, Heat Transfer in Condensation and Boiling, ISBN 0-387-52203-4, Springer Verlag, 1992.

[14] J. Chen, Correlation for boiling heat transfer to saturated liquids in convective ﬂow, Ind. Engng. Chem. Prog. Des. Dev. 5 (1966) 322–329.

[15] W. Rohsenow, A method for correlating heat transfer data for surface boiling of liquids, Trans. ASME, J. Heat Transfer 74 (1952) 969–976.

(13)

[17] S. Kutateladze, Elements of hydrodynamics of gas–liquid systems, Fluid Mech.––Sov. Res. 1 (1972) 29–50. [18] C. Tien, K. Chung, Entrainment limits in heat pipes, in: Proceedings of 3rd International Heat Pipe Conference,

Palo Alto (CA), 1978, pp. 36–40.

[19] M. Groll, S. R€oosler, Operation principles and performance of heat pipes and closed two-phase thermosyphons, J. Non-Equil. Thermody. 17 (1992) 91–151.